A better prognosis for genetic association studies in mice.

(A) A diagrammatic representation of the pattern of genetic variation within a region of the mouse genome. Each of the 18 identified SNPs within this genomic region is represented as a row, and the blue and yellow colored boxes indicate different alleles for each of the 12 strains analyzed. (B) The computational method for haplotype block formation will organize these SNP alleles into two haplotype blocks that accurately represent the pattern of genetic variation within this region. The first haplotype block has three different strain groupings (haplotypes), while the 2nd block has two different haplotypes. Strains 1–6, strains 7–9 and strains 10–12 have distinct genetic patterns, which gives rise to the three different haplotypes in the first haplotype block. While in the 2nd block, strains 1–3 and 7–8 have one allelic pattern, and strains 4–7 and 10–12 have the other allelic pattern, which produce the two haplotypes in this block. (C) In contrast, if the genetic variation is represented by the alleles at a single SNP-without knowing the true pattern of genetic variation within this region the allelic pattern and strain groupings will vary, depending upon the marker SNP that is selected to represent this region.

The difference in perspective when pulmonary H2-Eα mRNA expression is analyzed by linkage analysis in 2 strains, or in 10 inbred strains using HBCGM or GWAS methodology. (A) Graph showing pulmonary H2-Eα gene expression as the natural logarithm of the average of 3 independent measurements for each of 10 analyzed strains. The strains are divided into 3 distinct groups with high, intermediate or low levels of H2-Eα expression, which are indicated by different colored bars. The expression level of the DBA/2 strain (highest) is 269-fold greater than that of 129Sv (lowest), and is 37-fold greater than that of C3H/He (intermediate); while the two strains (C3H/He and C57BL6) analyzed by linkage analysis in [3] differ by only 4.5-fold (reproduced [7] with permission from Science). (B) HBCGM identified a haplotype block with 8 SNPs within the H2-Eα gene. Each SNP is represented as a row, and the colored boxes indicate the allele for each of the 10 analyzed strains. There were 3 different haplotypes within this region, and the one-way ANOVA model [7] showed a very strong correlation (p=1×10−7) between the haplotypic strain groupings and H2-Eα mRNA expression. (C) In contrast, a GWAS uses a two-sample t-test to assess the correlation between alleles at one SNP and the phenotypic data. The poor (p=0.054) or insignificant (p=0.31) correlation between individual selected SNP alleles (indicated by arrows) and the H2-Eα expression makes it impossible to detect the causal genetic locus if other spurious loci with smaller p-values were produced by the analysis, or after correction for multiple comparisons. This demonstrates the advantage that HBCGM has over GWAS methodology when multiple SNPs exert a composite effect on the phenotype.